MUSIC DATA ANALYSIS¶

Importing the required libraries for data analysis¶

In [1]:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
%matplotlib inline

Uploading the music file which is in form of csv¶

In [2]:
data = pd.read_csv("spotifyData.csv")

Checking the first 10 rows of the data¶

In [3]:
data.head()
Out[3]:
Artist Track Album Album_type Danceability Energy Loudness Speechiness Acousticness Instrumentalness ... Title Channel Views Likes Comments Licensed official_video Stream EnergyLiveness most_playedon
0 Gorillaz Feel Good Inc. Demon Days album 0.818 0.705 -6.679 0.1770 0.008360 0.002330 ... Gorillaz - Feel Good Inc. (Official Video) Gorillaz 693555221.0 6220896.0 169907.0 True True 1.040235e+09 1.150082 Spotify
1 Gorillaz Rhinestone Eyes Plastic Beach album 0.676 0.703 -5.815 0.0302 0.086900 0.000687 ... Gorillaz - Rhinestone Eyes [Storyboard Film] (... Gorillaz 72011645.0 1079128.0 31003.0 True True 3.100837e+08 15.183585 Spotify
2 Gorillaz New Gold (feat. Tame Impala and Bootie Brown) New Gold (feat. Tame Impala and Bootie Brown) single 0.695 0.923 -3.930 0.0522 0.042500 0.046900 ... Gorillaz - New Gold ft. Tame Impala & Bootie B... Gorillaz 8435055.0 282142.0 7399.0 True True 6.306347e+07 7.956897 Spotify
3 Gorillaz On Melancholy Hill Plastic Beach album 0.689 0.739 -5.810 0.0260 0.000015 0.509000 ... Gorillaz - On Melancholy Hill (Official Video) Gorillaz 211754952.0 1788577.0 55229.0 True True 4.346636e+08 11.546875 Spotify
4 Gorillaz Clint Eastwood Gorillaz album 0.663 0.694 -8.627 0.1710 0.025300 0.000000 ... Gorillaz - Clint Eastwood (Official Video) Gorillaz 618480958.0 6197318.0 155930.0 True True 6.172597e+08 9.942693 Youtube

5 rows × 24 columns

Our data consists of both qualitative and quantitative data

Checking numbers of rows and column of the data¶

In [4]:
data.shape
Out[4]:
(20594, 24)

The data has more than 10,000 rows and 24 columns

Checking the statistics of the quantitative variables in the data¶

In [5]:
data.describe()
Out[5]:
Danceability Energy Loudness Speechiness Acousticness Instrumentalness Liveness Valence Tempo Duration_min Views Likes Comments Stream EnergyLiveness
count 20594.000000 20594.000000 20594.000000 20594.000000 20594.000000 20594.000000 20594.000000 20594.000000 20594.000000 20594.000000 2.059400e+04 2.059400e+04 2.059400e+04 2.059400e+04 20592.000000
mean 0.620102 0.635176 -7.678254 0.096733 0.291391 0.056162 0.193653 0.530077 120.562616 3.742439 9.203740e+07 6.479902e+05 2.684679e+04 1.326446e+08 5.167227
std 0.165504 0.214274 4.639481 0.112182 0.286117 0.193622 0.168832 0.245542 29.588093 2.085211 2.726026e+08 1.773648e+06 1.911751e+05 2.423582e+08 4.117431
min 0.000000 0.000000 -46.251000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000049
25% 0.519000 0.507000 -8.868000 0.035700 0.045200 0.000000 0.094100 0.340000 96.994000 2.996746 1.478284e+06 1.754200e+04 4.060000e+02 1.559098e+07 2.386190
50% 0.638000 0.666000 -6.540500 0.050650 0.193000 0.000002 0.125000 0.538000 119.959000 3.551267 1.331348e+07 1.153155e+05 3.006000e+03 4.730525e+07 4.256881
75% 0.741000 0.798000 -4.935000 0.104000 0.476750 0.000474 0.237000 0.727000 139.923500 4.202163 6.739682e+07 5.000198e+05 1.373675e+04 1.343453e+08 6.822034
max 0.975000 1.000000 0.920000 0.964000 0.996000 1.000000 1.000000 0.993000 243.372000 77.934300 8.079649e+09 5.078865e+07 1.608314e+07 3.386520e+09 59.113924

A track with the mean, maximum and minimum views were 92037400, 8079649000 and 0 views,

Checking the data type of each column¶

In [6]:
data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 20594 entries, 0 to 20593
Data columns (total 24 columns):
 #   Column            Non-Null Count  Dtype  
---  ------            --------------  -----  
 0   Artist            20594 non-null  object 
 1   Track             20594 non-null  object 
 2   Album             20594 non-null  object 
 3   Album_type        20594 non-null  object 
 4   Danceability      20594 non-null  float64
 5   Energy            20594 non-null  float64
 6   Loudness          20594 non-null  float64
 7   Speechiness       20594 non-null  float64
 8   Acousticness      20594 non-null  float64
 9   Instrumentalness  20594 non-null  float64
 10  Liveness          20594 non-null  float64
 11  Valence           20594 non-null  float64
 12  Tempo             20594 non-null  float64
 13  Duration_min      20594 non-null  float64
 14  Title             20594 non-null  object 
 15  Channel           20594 non-null  object 
 16  Views             20594 non-null  float64
 17  Likes             20594 non-null  float64
 18  Comments          20594 non-null  float64
 19  Licensed          20594 non-null  object 
 20  official_video    20594 non-null  object 
 21  Stream            20594 non-null  float64
 22  EnergyLiveness    20592 non-null  float64
 23  most_playedon     20594 non-null  object 
dtypes: float64(15), object(9)
memory usage: 3.8+ MB

Our data has both qualitative data(Artist,Album,Album) and quantitative data

Data Cleaning¶

Checking if there is any null value in our dataset

In [7]:
data.isnull().sum()
Out[7]:
Artist              0
Track               0
Album               0
Album_type          0
Danceability        0
Energy              0
Loudness            0
Speechiness         0
Acousticness        0
Instrumentalness    0
Liveness            0
Valence             0
Tempo               0
Duration_min        0
Title               0
Channel             0
Views               0
Likes               0
Comments            0
Licensed            0
official_video      0
Stream              0
EnergyLiveness      2
most_playedon       0
dtype: int64

Lets drop the null values in the column EnergyLiveness

In [8]:
data = data.dropna()
In [9]:
data.isnull().sum()
Out[9]:
Artist              0
Track               0
Album               0
Album_type          0
Danceability        0
Energy              0
Loudness            0
Speechiness         0
Acousticness        0
Instrumentalness    0
Liveness            0
Valence             0
Tempo               0
Duration_min        0
Title               0
Channel             0
Views               0
Likes               0
Comments            0
Licensed            0
official_video      0
Stream              0
EnergyLiveness      0
most_playedon       0
dtype: int64

The null value still exists.So lets convert the value that represents the null value to NaN so that it can be easy to cleaned¶

In [10]:
missingValue = ["??","na","nan","x","?",""]
data = data.replace(missingValue,np.NaN)
data = data.dropna()
In [11]:
data.isnull().sum()
Out[11]:
Artist              0
Track               0
Album               0
Album_type          0
Danceability        0
Energy              0
Loudness            0
Speechiness         0
Acousticness        0
Instrumentalness    0
Liveness            0
Valence             0
Tempo               0
Duration_min        0
Title               0
Channel             0
Views               0
Likes               0
Comments            0
Licensed            0
official_video      0
Stream              0
EnergyLiveness      0
most_playedon       0
dtype: int64
Now we have cleaned our dataset by eliminating all the null values¶

Data Visualization¶

Lets see the correlation between quantitative variables¶

In [12]:
corr = data[["Danceability","Energy","Loudness","Speechiness","Acousticness","Instrumentalness","Liveness","Valence","Likes","Views","Stream","Comments"]].corr()
sns.set(style = "darkgrid")
plt.figure(figsize = (12,8))
sns.heatmap(corr,annot = True,fmt = ".0%")
Out[12]:
<Axes: >

There is a high positive correlation between Energy and Loudness,Likes and Views,Likes and Streams,Likes and comments and Views and Streams, also there is a high negative correlation between Acousticness and Energy,Acousticness and Loudness,and Instruments and Loudness¶

Lets view the positive correlation between Energy and Loudness¶

In [13]:
sns.scatterplot(x= "Energy",y="Loudness",data =data)
plt.title("Correlation between Energy and Loudness")
plt.show()

The higer the loudness the higher the energy¶

Lets view the negative correlation between Energy and Acousticness¶

In [14]:
sns.scatterplot(x= "Energy",y="Acousticness",data =data)
plt.title("Correlation between Energy and Acousticness")
plt.show()

While Energy increases, Acousticness level decreases¶

Lets view the distribution plot¶

In [15]:
distplot = data[["Danceability","Energy","Loudness","Speechiness","Acousticness","Instrumentalness","Liveness","Valence","Likes","Views","Stream","Comments"]]
sns.pairplot(distplot)
Out[15]:
<seaborn.axisgrid.PairGrid at 0x7fce6d0e7610>

Lets compare which platform(between Youtube and Spotify) had the most played songs¶

In [16]:
plt.pie(data["most_playedon"].value_counts(),labels = data["most_playedon"].value_counts().index,autopct = "%.2f%%")
plt.title("Count of most_playedon for both Youtube and Spotify")
plt.show()
In [17]:
ax= sns.countplot(x= "most_playedon",data = data,palette = "bright")
for label in ax.containers:
    ax.bar_label(label)
plt.title("Count of most_playedon for both Youtube and Spotify",fontsize =13)
plt.show()
Most tracks were played on spotify which had 76.20% tracks while youtube had only 23.80% tracks. It seems that most people prefer to listen music through Spotify platform compared to youtube¶

Lets compare which platform played the highest number of Licenced music¶

In [18]:
ax = sns.countplot(x = "most_playedon",hue = "Licensed",palette = "bright",data = data )
for label in ax.containers:
    ax.bar_label(label)
plt.title("Count of licensed an not licensed tracks from each platform",fontsize = 14)
plt.show()
The platform with the most licensed tracks was youtube which is around 82% of its total compared to spotify which had only around 63% of its track licensed¶

Lets see who were the top 10 artist with the highest Views with respect to the Energy and Acousticness¶

In [19]:
data[["Artist","Energy","Acousticness","Views"]].groupby("Artist").mean().sort_values(by="Views",ascending = False)[:10]
Out[19]:
Energy Acousticness Views
Artist
Ed Sheeran 0.6089 0.308300 1.546021e+09
CoComelon 0.4002 0.529600 1.460167e+09
Katy Perry 0.7655 0.041185 1.312063e+09
Charlie Puth 0.5650 0.381210 1.216759e+09
Luis Fonsi 0.7417 0.314690 1.162811e+09
Justin Bieber 0.6003 0.373010 1.099106e+09
Daddy Yankee 0.8212 0.133920 1.087193e+09
Bruno Mars 0.6256 0.197940 1.024092e+09
Macklemore & Ryan Lewis 0.6699 0.223420 1.012206e+09
Coldplay 0.5704 0.207006 9.997278e+08
In [20]:
z = data[["Artist","Energy","Acousticness","Views"]].groupby("Artist").mean().sort_values(by="Views",ascending = False)[:10]
In [21]:
plt.figure(figsize = (20,10))
ax = sns.barplot(x=z.index,y= "Views",data =z)
for label in ax.containers:
    ax.bar_label(label)
plt.title("Top 10 Artist with respect to their views",fontsize = 20)
plt.show()

Lets see top 10 artists who had the highest Streams with respect to Energy and Acousticness¶

In [22]:
data[["Artist","Energy","Acousticness","Stream"]].groupby("Artist").mean().sort_values(by="Stream",ascending = False)[:10]
Out[22]:
Energy Acousticness Stream
Artist
Post Malone 0.6350 0.221320 1.525126e+09
Ed Sheeran 0.6089 0.308300 1.439488e+09
Dua Lipa 0.7347 0.045934 1.340808e+09
The Weeknd 0.6425 0.124516 1.303197e+09
XXXTENTACION 0.5075 0.398617 1.233362e+09
Justin Bieber 0.6003 0.373010 1.209777e+09
Imagine Dragons 0.7174 0.095532 1.185831e+09
Coldplay 0.5704 0.207006 1.177848e+09
Khalid 0.5937 0.291010 1.138684e+09
Bruno Mars 0.6256 0.197940 1.089786e+09
Ed Sheeran leads with the highest Views while Post Malone Leads with the highest stream.¶

In either Views or Stream,Energy of both top 10 had high energy to be successful¶

Lets now compare top 15 tracks with the highest views with relation to energy and Acousticness and know the Artist of the Track¶

In [23]:
top15 = data[["Artist","Track","Energy","Valence","Stream","Views"]].groupby(["Track","Artist"]).mean().sort_values(by="Views",ascending = False)[:15]
data[["Artist","Track","Energy","Valence","Stream","Views"]].groupby(["Track","Artist"]).mean().sort_values(by="Views",ascending = False)[:15]
Out[23]:
Energy Valence Stream Views
Track Artist
Despacito Luis Fonsi 0.797 0.839 1.506598e+09 8.079649e+09
Daddy Yankee 0.797 0.839 1.506598e+09 8.079647e+09
Shape of You Ed Sheeran 0.652 0.931 3.362005e+09 5.908398e+09
See You Again (feat. Charlie Puth) Charlie Puth 0.481 0.283 1.521255e+09 5.773798e+09
Wiz Khalifa 0.481 0.283 1.521255e+09 5.773797e+09
Wheels on the Bus CoComelon 0.387 0.965 8.343436e+07 4.898831e+09
Uptown Funk (feat. Bruno Mars) Mark Ronson 0.609 0.928 1.653820e+09 4.821016e+09
Gangnam Style (?????) PSY 0.937 0.749 3.709911e+08 4.679767e+09
Sugar Maroon 5 0.788 0.884 1.502781e+09 3.817733e+09
Roar Katy Perry 0.771 0.436 8.847210e+08 3.725749e+09
Counting Stars OneRepublic 0.705 0.477 1.805320e+09 3.721610e+09
Sorry Justin Bieber 0.760 0.410 1.740759e+09 3.627306e+09
Thinking out Loud Ed Sheeran 0.445 0.591 2.154334e+09 3.547156e+09
Baa Baa Black Sheep CoComelon 0.396 0.969 4.655435e+07 3.486504e+09
Waka Waka (This Time for Africa) [The Official 2010 FIFA World Cup (TM) Song] (feat. Freshlyground) Shakira 0.871 0.753 6.299185e+08 3.463816e+09
In [24]:
top10 =  data[["Track","Energy","Valence","Stream","Views"]].groupby(["Track"]).mean().sort_values(by="Views",ascending = False)[:10]
plt.figure(figsize = (27,20))
ax = sns.barplot(x=top10.index,y= "Views",data =top10)
for label in ax.containers:
    ax.bar_label(label)
plt.title("Top 10 Tracks with respect to their views",fontsize = 30)
plt.show()
Ed Sheeran and Cocomelon had 2 tracks which at top 15 in Views¶

Lets rank the top 15 Artists with regard to positivity in their tracks¶

In [25]:
topValenceTrack=data[["Artist","Energy","Valence","Stream","Views"]].groupby(["Artist"]).mean().sort_values(by="Views",ascending = False)[:15]
topValenceTrack=topValenceTrack.sort_values(by = "Valence",ascending=False)
topValenceTrack
Out[25]:
Energy Valence Stream Views
Artist
CoComelon 0.4002 0.82380 3.690822e+07 1.460167e+09
Daddy Yankee 0.8212 0.68220 5.135462e+08 1.087193e+09
Luis Fonsi 0.7417 0.64480 5.054326e+08 1.162811e+09
Calvin Harris 0.8579 0.63190 9.548542e+08 9.758476e+08
Charlie Puth 0.5650 0.61090 7.041187e+08 1.216759e+09
Katy Perry 0.7655 0.60140 6.607329e+08 1.312063e+09
Bruno Mars 0.6256 0.59974 1.089786e+09 1.024092e+09
Ed Sheeran 0.6089 0.56600 1.439488e+09 1.546021e+09
Justin Bieber 0.6003 0.56130 1.209777e+09 1.099106e+09
BLACKPINK 0.7681 0.53370 4.015972e+08 9.392962e+08
Eminem 0.7826 0.51860 9.748588e+08 9.119324e+08
Macklemore & Ryan Lewis 0.6699 0.50833 3.982978e+08 1.012206e+09
Imagine Dragons 0.7174 0.45197 1.185831e+09 9.093785e+08
DJ Snake 0.7901 0.30885 7.739463e+08 9.180268e+08
Coldplay 0.5704 0.29060 1.177848e+09 9.997278e+08
In [27]:
plt.figure(figsize = (25,10))
ax = sns.barplot(x=topValenceTrack.index,y= "Valence",data =topValenceTrack)
for label in ax.containers:
    ax.bar_label(label)
plt.title("Top 10 Tracks with respect to their Positive Energy",fontsize = 30)
plt.show()
Cocomelon who is ranked top 10 in Views has tracks that are positive¶
80% of the tracks which are top 15 Views in category have positive energy¶

Lets see the song with the highest energy¶

In [28]:
topSongEnergy = data[["Track","Energy","Valence","Stream","Views"]].groupby(["Track"]).mean().sort_values(by="Views",ascending = False)[:10]
topSongEnergy=topSongEnergy.sort_values(by="Energy",ascending = False)
In [29]:
plt.figure(figsize = (25,10))
ax = sns.barplot(x=topSongEnergy.index,y= "Energy",data =topSongEnergy)
for label in ax.containers:
    ax.bar_label(label)
plt.title("Top 10 Tracks with respect to Energy",fontsize = 30)
plt.show()
"Gangnam Style" by PSY had the highest Energy in the Ranking while "Wheels on the Bus" by Cocomelon was the list of the top 10 songs¶

Challenge¶

Our data does not have Year column which could help in finding out type of track and the year it had the highest views,or even compare Artists popularity with regards to the year he or she was famous¶

Results and Conclusion¶

Artist who produce music which have high Energy,high Valence(Positivity),and less Acousticness such as Edsheeran and Cocomelon tend to have higher views and the tracks are highly streamed,and the platform mostly used by people to listen to music is Spotify compared to Youtube.¶
Artists should consider the above mentioned factors to optimize on their success¶
In [ ]: